On Thursday morning, the Japanese Ministry of Finance released its weekly International Transactions in Securities data covering March 1 to 7, the first full week of the Iran war. Within hours, a widely shared post on X announced: the BOJ had sold ¥400 billion in foreign bonds. The carry trade was still unwinding into the oil shock.

This reading is wrong on every count.

The data in question comes from Section 1 of the MOF release: Portfolio Investment Assets, which tracks transactions by Japanese residents, not the Bank of Japan. The BOJ’s foreign reserve operations are excluded from this dataset entirely. The ¥400 billion figure was a net purchase of foreign long-term debt securities by Japanese institutional investors. Positive in Section 1 means net acquisition, not net disposition. The sign convention changed in January 2014: plus now denotes buying, not selling.

Three claims circulated. All three were wrong.

The first claim was that the BOJ sold foreign bonds. It did not. The “residents” in Section 1 are designated major investors: banks, securities firms, life insurers, and asset managers. The BOJ is not among them.

The second claim was that ¥400 billion of foreign bonds were sold. The opposite occurred. Japanese institutions bought ¥400 billion net in foreign long-term debt, the first outward deployment in five weeks. This reversed the repatriation trend that had prevailed since late January. Capital flowed out of Japan, not in.

The third claim was that the carry trade was unwinding. Carry trade unwinding requires selling foreign assets, converting proceeds to yen, and driving yen appreciation. What the data showed was the reverse: Japanese institutions deployed capital abroad, buying foreign bonds at crisis-elevated yields. This is carry-like behaviour. It is yen-negative, not yen-positive.

The correct reading of the full dataset tells a more nuanced story than any single line.

Foreign equity buying of Japanese stocks came in at approximately ¥386 billion net. This extends the streak of net foreign buying to 11 consecutive weeks. The streak survived the first week of war, but the flow halved from the ¥974 billion recorded the prior week. Wounded, not dead.

Foreign selling of Japanese bonds was the real de-risking event. Non-residents sold approximately ¥964 billion in Japanese debt, a massive reversal from the ¥1.37 trillion of net buying the week before. This was the first net selling in three weeks. The flight from Japan happened in fixed income, not equities.

Japanese outward bond investment of ¥400 billion net reversed five weeks of repatriation. Japanese institutions used the crisis to buy foreign bonds, deploying abroad rather than pulling money home.

The narrative on X said: foreigners fleeing Japan, carry trade unwinding, BOJ selling bonds. The data said: foreigners still buying equities (at reduced pace), foreigners dumping Japanese bonds (genuine de-risking), and Japanese institutions buying foreign bonds (deploying abroad). Three out of three narratives were wrong or misleading.

Why does this happen? The MOF data is genuinely difficult to read. The weekly release is a PDF published in both Japanese and English, but the formatting is dense and the sign conventions are counterintuitive for anyone who started reading the series before 2014. Section 1 covers resident outward investment. Section 2 covers non-resident inward investment. Confusing the two is easy. Getting the sign backwards is easier still.

The deeper problem is structural. In a crisis, X amplifies narratives faster than anyone can check the data. The carry-trade-unwinding story is emotionally compelling: it connects to the August 2024 yen carry unwind that remains seared into market memory, it sounds sophisticated, and it generates engagement. The actual MOF data is boring, published at 08:50 JST in a format that discourages casual reading, and requires familiarity with reporting conventions that even experienced analysts occasionally misread.

The gap between what X says and what the data says is where analytical edge lives. Anyone can read X. Reading the MOF CSV gives you the real story.

A few practical notes for anyone who wants to check this themselves. The MOF weekly data page publishes the latest release as a PDF. The FAQ explains the reporting conventions. The critical detail: from January 2014, net purchases are shown with a plus sign and net sales with a minus sign. Before that date, the convention was reversed. If you are looking at historical data that spans 2014, adjust accordingly. The data covers designated major investors only, not the full universe of Japanese financial institutions, and it does not include the BOJ.

None of this is secret. It is just tedious. In a market that rewards speed over accuracy, tedium is edge.

— Gyokuro

The views expressed here are the author’s own and do not constitute investment advice. This blog is independent, carries no affiliate links, and receives no compensation from any entity mentioned.